It’s a race to the bottom, as online social networks trip over themselves to collect ever greater amounts of data about their users. With sentiment towards Facebook at an all-time-low, it’s high-time for social networks to rethink their stance on online privacy, lest they permanently alienate their user base, or worse, get themselves legislated to oblivion.
The key motivator that drives social networks to collect personal information is advertising. The theory is that by crunching all of the digital traces of an individual, advertisers can reach serendipity: the ability to predict what people want and sell it to them before they even know they want it. The well publicized downside of this approach is that people find it Orwellian, and the backlash from the media has been fierce.
There is another problem with ad targeting based on social information: it doesn’t work nearly as well as advertisers wish it would. As sociologists like Danah Boyd point out, online personalities are fundamentally disconnected from real world identity. This makes their utility as predictors for ad targeting highly suspect. At best, online social profiles are just a facet of who we truly are, due to the inherent constraints on expressiveness of online media. At worst, online identities are total fabrications, engineered to appeal to massive online crowds, totally divorced from reality.
Research by people like network scientist Duncan Watts (whose new book I highly recommend) demonstrates that extrapolating the future actions of humans based on past observations suffers from diminishing returns. On a hypothetical percentile scale, where 0% is totally wrong prediction, and 100% is omniscience, basic information like age, gender, marital status, and income bracket allow for predictions of roughly 50% accuracy. More specific information on personal interests, such as favorite books and movies, maybe pushes predictive accuracy up to 60%. Piling on yet more information, like keywords from a recent conversation on Facebook, may push accuracy up to 65%. The point is that the vast majority of predictive accuracy comes from only the most basic information; more detailed, and thus more invasive, information only increases accuracy by a fractional amount.
Social networks need to reevaluate the cost benefit analysis of their data collection policies. As Google has shown, most people will tolerate some level of online tracking in exchange for free service, and it has made them wildly successful. On the other hand, Facebook, despite their vast troves of data, makes six-times less per user than Google. What is the point of abusing people’s expectations of online privacy if it fundamentally cannot make you more money, and only serves to disenfranchise your users?
Image: Metro Centric